An Examination of the Crystallization of Urea from Supersaturated

---

An Examination of the Crystallization of Urea from Supersaturated Aqueous and Aqueous-Methanol Solutions as Monitored In-Process Using ATR FTIR Spectroscopy

CRYSTAL GROWTH & DESIGN 2004 VOL. 4, NO. 5 929-936

Heidi Groen† Centre for Molecular and Interface Engineering, School of Engineering and Physical Sciences, Heriot-Watt University, Edinburgh EH14 4AS, UK

Kevin J. Roberts* Institute of Particle Science and Engineering, School of Process, Environmental and Materials Engineering, University of Leeds, Leeds LS2 9JT, UK Received November 3, 2003;

Revised Manuscript Received May 20, 2004

ABSTRACT: In-process ATR FTIR spectroscopy combined with optical turbidometry is applied to the measurement of reactant supersaturation during temperature-programmed studies of the batch crystallization of urea from aqueous and mixed aqueous/methanolic solvent systems at the 400 mL scale size (pitched blade impeller) for solute compositions in the 35-65 w/w% range. The metastable zone widths were found to be in the range 3-12 °C and critical supersaturations in the 1.05-1.18 range, respectively, inversely dependent upon solute concentration. Only a slight dependency of the metastable zone width with reactor heating/cooling rates, indicative of fairly matched rates of nucleation and solubility change, is found. Cross-correlation between FTIR and turbidometic measurements reveal, as expected, the former technique to be more sensitive to the detection of the crystallization onset. No evidence was found for phase separation (oiling-out) prior to nucleation, in contrast to previous studies of citric acid (Groen, H.; Roberts, K. J. J. Phys. Chem. B 2001, 105, 10723-10730). Crystallization enthalpic release at the onset point is found to affect transient crystal dissolution and regrowth following nucleation. Good FTIR calibration, with excellent correlation to known solubility data, using a univariant IR transmittance (peak intensity) model is achieved for both single and mixed solvent systems, this despite slight variation in the IR peak positions with temperature and concentration. Attempts to use the calibration data at higher solute concentrations beyond the calibration range were not successful, this being associated with a significant underestimation of the solution supersaturation. A forward look to improvements in FTIR calibration using multivariant chemometric approaches is highlighted. 1. Introduction Recent years have seen a significant increase in interest in the development and use of process analytical techniques (PAT) for the monitoring of crystallization processes (e.g., refs 2 and 3). Within this reaction, supersaturation monitoring is particularly important in industrial crystallization processes as it provides a potential method for controlling crystal size and its distribution. The use of attenuated total reflectance (ATR) FTIR spectroscopy to determine the level of supersaturation during crystallization experiments was proposed by Dunuwila et al.4 First measurements conducted using a Micro Circle open boat cell equipped with a ZnSe ATR rod were carried out on aqueous citric acid solutions. These experiments demonstrated the feasibility of this technique for quantitative determination of the concentration of solution species and, via knowledge of solubility, its supersaturation. This work and more recent studies have confined the utility of ATR FTIR spectroscopy for practical supersaturation monitoring5-13 as well as its further development in supersaturation control12 and for crystallization process design when used in combination with other PAT * To whom correspondence should be addressed. E-mail: [email protected]. † Present address: Degussa AG, Process Technology-New Processes, 63457 Hanau, Germany; e-mail: [email protected].

methodology.13 Overall, such developments have value in the design of optimal process routes for the manufacture of high value-added organic fine chemical products such as pharmaceuticals as well as facilitating potential approaches for improvement in product quality. In previous work, in-process FTIR spectroscopy using an ATR probe has been successfully applied to determine the solubility, supersolubility, and phase diagram of citric acid anhydrate and monohydrate in aqueous solution.1 In this, ATR FTIR spectroscopy, used in combination with optical turbidity measurements, revealed the occurrences of solution phase separation (oiling out) prior to nucleation. The extension of the ATR FTIR technique to the monitoring and controlling of more complicated systems such as mixed solvent solutions is important in view of its potential application to the control of the batch crystallization from reaction mother liquor for organic compounds following chemical synthesis. Such systems offer some challenge for subsequent analysis in that the concentration dependency of the solute in mixed solvents can be rather nonideal. Hence, this paper focuses on a study of crystallization from a mixed solvent system in which calibrated inprocess ATR FTIR measurements have been used to monitor the crystallization process. The system chosen for study was urea crystallizing from aqueous and mixed

10.1021/cg030038y CCC: $27.50 © 2004 American Chemical Society Published on Web 08/11/2004

930

Crystal Growth & Design, Vol. 4, No. 5, 2004

Groen and Roberts

water/methanol solutions. Analysis of such a system is important in that urea is a highly soluble material in aqueous solution which crystallizes in the laboratory comparatively easily and which would be expected to have a rather small metastable zone width (MSZW). Thus, it exhibits completely different crystallization behavior in comparison to citric acid,4 which displays the opposite properties. In addition, this solute/solvent system has not been previously characterized in terms of the concentration dependency of its IR spectra. Reflecting this, initial experiments were conducted with urea in aqueous solutions. The feasibility of extrapolating and predicting IR calibration curves into concentration regions beyond the experimental data calibration range was also examined. 2. Materials and Methods 2.1. Materials. Aqueous solutions of different concentrations of urea were prepared with distilled water. Watermethanol solutions, consisting of 80% weight/weight (w/w) methanol and 20% (w/w) water, were prepared with distilled water and HPLC-grade methanol. Methanol (formula: CH3OH; molecular weight: 32.04) and urea (formula: H2NCONH2; molecular weight: 60.06) were purchased from the Aldrich Chemical Co. 2.2. Experimental Setup. In-situ measurements of solution concentration were carried out using a Dipper-210 ATR FTIR immersion probe equipped with a ZnSe conical internal reflection element manufactured by Axiom Analytical Incorporated together with a Bomen WorkIR FTIR spectrometer connected to a PC equipped with Grams software (Galactic Industries Corporation). Experiments were carried out using a 400-mL jacketed glass crystallizer which could be rapidly cooled via the use of a pair of Haake F3 circulating water baths that could be set to different temperatures and switched between via computer controlled valves. Stirring was provided using a pitched blade glass stirrer at a constant speed of 330 rpm. Temperature and turbidity of the solution/slurry was measured using a PT100 thermometer and a Sybron Brinkmann Lexan fiber optic probe linked with a Brinkmann PC700 colorimeter, respectively. The turbidity sensor was calibrated to be 100% for a clear solution and 0% for the cloudy (crystal/solution slurry) liquid postcrystallization. All experimental data was logged using a computer-based data acquisition system. 2.3. ATR FTIR Calibration Measurements. To determine the concentration, solubility, and degree of supersaturation of urea in the respective solvent, calibration curves of a specific transmittance ratio versus concentration of urea in water were developed as follows. Appropriate amounts of urea and solvent were heated by a Haake F3 circulating water bath and stirred to a homogeneous solution in the respective crystallizer. Three ATR FTIR spectra (16 scans/spectrum) of the solution in intervals of 15 min were scanned at thermal equilibrium at set temperatures of 10, 30, 50, and 70 °C. The solubility curves of urea in water and in the mixed methanol-water solvent were determined over a wide temperature range from 10 to 70 °C using the ATR FTIR transmittance ratios of slurries at equilibrium. Slurries of urea in the respective solvent were prepared in the batch crystallizer. An excess amount of urea was stirred in an appropriate amount of solvent for 6 h at 10 °C before the ATR FTIR spectra were scanned (16 scans/spectrum). It was checked that after 6 h the ATR FTIR spectrum would not show any concentration changes at all, i.e., thermodynamic equilibrium was reached. Three spectra were scanned of the slurries at equilibrium per setting in 15-min time intervals. After that, the temperature was increased step by step by 10 °C and the procedure was repeated. The temperature settings ranged from 10 to 70 °C. 2.4. Crystallization Protocol and Associated In-Process Measurements. For studies of aqueous solutions, solute

Figure 1. Average dissolution and crystallization temperatures of 140 g (58.3 w/w) and 180 g (64.3% w/w) of urea in 100 g of water as a function of the cooling rate. concentration of 140 and 180 g, per 100 g of water, respectively, 58.3 and 64.3% (w/w), were examined. For studies of the mixed methanol-water (MW) solution 519, 939, and 695 g per 100 g of solvent, respectively, 32.4, 37.2, and 39.6% w/w, were examined. The crystallization experiments were carried out at three different linear cooling and heating rates 0.1, 0.25, and 0.33 °C/min with each experiment being repeated three times each to ensure consistency. By examination of the onset points for crystallization (Tcryst) and dissolution (Tdiss) obtained from the turbidity probe during cooling and heating experiments, respectively, the MSZW for the urea solutions was measured as a function of heating/cooling rates. Extrapolation of the data to zero rate was carried out to determine Tcryst, Tdiss, and the MSZW, independent of the applied cooling rate (Figure 1). ATR FTIR spectra (16 scans/spectrum) were accumulated every 30 s during crystallization and dissolution processes and cross-correlated with the measured temperature/turbidity data.

3. Results and Discussion 3.1. Crystallization of Urea from Aqueous Solution. 3.1.1. Metastable Zone Width Determination. The results of the MSZW measurements are given in Figure 1 revealing the MSZW, for the virtual cooling rate of 0 K/min, to be smaller for the higher concentration of urea in water (3.82 K) compared the lower concentration of urea in water (7.28 K). It is noteworthy that the cooling rate dependence of the crystallization process is rather low. This reflects the fact that the nucleation rate is probably only slightly lower than the rate of solute concentration decrease over this range of cooling rates. From this, it can be concluded that urea, in general, appears to nucleate quite easily from aqueous solution with the MSZW being fairly small, albeit increasing with solute concentration and depending slightly on the cooling/heating rates used. 3.1.2. ATR FTIR Spectra of Aqueous Urea Solutions Measured as a Function of Solute Concentration. Representative ATR FTIR spectra of aqueous urea solutions are shown in Figure 2. The spectra show three urea bands that clearly change in intensity. The broad band from 3100 to 3500 cm-1 consists of overlapping water and amine bands. It was found that the solvent band at 3279 cm-1 shows a systematic urea concentration dependency, so this band and two potential solute bands at 1462 cm-1

Crystallization of Urea

Crystal Growth & Design, Vol. 4, No. 5, 2004 931

Figure 2. ATR FTIR spectra of aqueous urea solutions compared to the spectrum of water. Spectra recorded here were taken at 50°C; spectra at 10, 30, and 70 °C follow the same trend. Table 1. Definition of IR Transmittance Ratios as Utilized for the Calibration of Aqueous Urea Solutions RT ) transmittance of water band at T1 cm-1/ transmittance of urea band at T2 cm-1 RT [-]

T1 [cm-1]

RT1′ RT1′′ RT2′′

3279 3279 3279

T2 [cm-1] min (∼1460) 1447 1160

(RT1) and 1160 cm-1 (RT2) were identified for peak ratioing (see Table 1). It was noticed that the peak maxima for the latter ranged between 1468 and 1446 cm-1 depending on temperature and concentration of the solution (Figure 3). Such a peak shift is not unexpected, and these shifts can be rationalized with changes in hydrogen bonding between the water and urea as concentration and temperature change14 as hydrogen bonds influence bond stiffness, which, in turn, alters the vibration frequency. From Figure 3a, it is clear that hydrogen bonding increases with increasing concentration of urea in water, hence the peak shift toward lower wavenumbers. Increasing temperature, in turn, means that each molecule will have more energy on average, and hence weak associate forces are likely to be broken. This would lead to a lesser degree of

hydrogen bonding, and thus changes in frequency to higher wavenumbers would be observed for the groups forming hydrogen bonds, as the -N-H bending mode here. Experimental data confirms this shift with temperature (Figure 3b). 3.1.3. ATR FTIR Calibration for Aqueous Urea Solutions. To investigate the influence of the peak shift on the calibration, one calibration parameter (RT1′) was calculated from the minimum transmittance within this range and another one (RT1′′) was calculated from a transmittance value at a set wavenumber that corresponds to the minimum transmittance of a saturated solution at 70 °C (1447 cm-1). This was chosen because at this wavenumber the peak intensity changes with concentration are most evident, and the minimum would not be crossed over at any time for any monitored concentration. The defined IR transmittance ratios are given in Table 1. The plot of RT1′′ versus concentration (Figure 4) was found to be consistent with two different fitting functions centered at about 30% (w/w) with both these sets of data, i.e., low and high urea concentrations obeying an exponential function. Reflecting this observation, the other calibration curves were also examined, and it was found that the same trend was followed (Figure 4b,c). This result suggests that the calibration function RT ) a exp(bcs) as used for the previous work on citric acid4 with a and b being functions of temperature, might be too simple, when having to deal with overlapping IR bands from the solute and the solvent (3100-3500 cm-1 - water and amine bands). In this case, the absorption of the solute i as well as the solvent has to be taken into account for the overall absorption across the IR band. This gives the following simplified relationship between RT and the concentration of the solute ci according to the Lambert-Beer Law:

RT )

Tsolv ) exp{2.3Ndp(ici - solvcsolv)} Ti

(1)

where N is the number of reflection points, dp is the depth of penetration, and  is the molar absorption coefficient with the factor 2.3 being due to conversion to the natural logarithm.

Figure 3. (a, b) ATR FTIR spectra of aqueous urea solutions, which clearly show the peak shift of the band at around 1460 cm-1: (a, left) toward lower wavenumbers as the concentration increases from 0% (w/w) to 60% (w/w) urea in water (in 10% (w/w) steps): (b, right) toward higher wavenumbers as the temperature increases from 10 to 70 °C (constant concentration of 40% (w/w) urea in water).

932

Crystal Growth & Design, Vol. 4, No. 5, 2004

Groen and Roberts

Figure 5. Solubility of urea in aqueous solution, comparing the by ATR FTIR experimentally determined values with literature values.16 The solubility curve is a least-squares fit of the ATR FTIR data.

Figure 4. Calibration of the transmittance ratios RT1′′ (a), RT1′ (b), and RT2′′ (c) versus the concentration of urea in aqueous solution.

From eq 1,1 it becomes apparent that for high concentrations of the solute i, the first term within the exponent becomes dominant, and for low concentrations of i, the second term becomes dominant. This would explain the experimental observation of the above calibration curves for RT1′, RT1′′, and RT2′′ exhibiting two exponential fitting curves hinging at a concentration of 30% (w/w) urea in water. To extract a threedimensional calibration grid for RT depending on the concentration ci and temperature, this behavior can be modeled as follows:

RT ) a exp(bci,low)

(2)

RT ) a exp(bci,high)

(3)

and

again with a and b being functions of temperature. Hence, it is very important to know in which concentration range the ATR FTIR method is applied to extract

supersaturation if erratic data are not to be produced through the use of calibration curves, which might only be valid for specific concentration ranges. It can be seen from Figure 4 that the best fit of the experimental data was achieved for the calibration parameter RT1′′, i.e., based on utilization of the fixed peak at 1447 cm-1. This calibration ratio reveals the temperature dependency of the band is much more clear than for RT1′, the determination of which is based on the minimum transmittance of the same peak as it shifts with concentration and temperature. Therefore, it is more straightforward to use RT1′′ for detection of concentration and supersaturation, as no time-consuming analysis to extract the minimum of the moving peak is necessary, and the temperature dependence is more significant and therefore easier to calibrate. From Figure 4c, it can be seen that the experimental data for RT2′′ exhibits the most scattered results, making it difficult to be utilized reliably for concentration measurement. This is probably due to the weaker intensity of the band used in its determination, resulting in a worse signal-to-noise ratio. 3.1.4. Solubility of Urea in Aqueous Solution. With the above calibration method, the solubility of aqueous urea solutions was determined by ATR FTIR measurements (Figure 5) and compared to literature values.15 The solubility data determined by ATR FTIR measurements is in pleasing agreement with the dissolution temperatures as determined by optical turbidometric measurements and literature data.15 It can also be seen from Figure 5 that there is no significant difference between the results calculated from the parameter using the minimum transmittance of the peak at around 1460 cm-1 (RT1′) and the one using the fixed wavenumber at 1447 cm-1 (RT1′′). This, once again, supports the functionality of RT1′′. 3.1.5. Supersaturation Monitoring during Urea Crystallization from Aqueous Solution. Crystallization processes of aqueous urea solutions of two different concentrations were monitored on-line by ATR FTIR spectroscopy. In this, the lower concentration (58.3% w/w) was chosen just within the developed calibration range, and the higher one (64.3% w/w) was chosen just above the calibration range and can only be extrapolated by the least-squares fitting curves (see Figure 4).

Crystallization of Urea

Figure 6. Change of RT1′′ during a temperature induced crystallization process. Also shown in the plot is the solubility of urea in water as measured by ATR FTIR. The solution transmittance is plotted to indicate the onset of turbidity and dissolution (concentration: 58.3% w/w in aqueous solution; cooling/heating rate: 0.25 K/min).

Figure 6 shows the change of RT1′′ during a temperature cooling-induced crystallization experiment of urea from aqueous solution. As can be seen by the sudden drop of RT1′′, urea crystallizes quite rapidly, the onset of which being slightly before the observed onset as measured via the turbidity probe. Interestingly, this shows, contrary to the results obtained from monitoring citric acid crystallization processes,1 that the onset of crystallization was detected earlier by ATR FTIR than by optical turbidity measurements. This nicely illustrates the enhanced sensitivity of the ATR FTIR technique compared to optical turbidometric sensing. This is not unexpected mindful of the fact that direct sensing of particles by light is bound to be limited to a crystal size range of ca. 0.5 µm. In contrast FTIR, via monitoring concentration change, should be more sensitive to detecting the crystallization onset point. Furthermore, these data support previous studies1 using combined ATR FTIR/turbidity in which the reverse process was observed, i.e., optical turbidity change prior to the depletion of citric acid concentration. This latter behavior was interpreted1 as solution phase separation due to incipient “oiling out” prior to crystallization. Close examination of both the ATR and turbidity data reveals a sudden temperature increase just after the crystallization onset point. Presumably, this reflects the release of the heat of crystallization, which in turn causes transient solution heating and associated partial crystal dissolution. However, this effect appears to be rather short-lived with the supersaturation being found subsequently to decay to equilibrium following further cooling causing the concentration to follow the saturation curve. During the reheating, the solution was found to become undersaturated (ATR FTIR data) simultaneously with the observation of dissolution onset (turbidity data). Figure 7 shows the change of the spectral ratio RT1′′ for the three different cooling rates during temperatureinduced crystallization experiments for the 58.3% w/w concentration solution, nicely illustrating the cooling rate dependency of the MSZW as recorded on-line by ATR FTIR spectroscopy. Using the urea/water calibration data, the raw ATR FTIR data was converted into concentration units and the resulting supersaturation profile during the crystallization and dissolution process

Crystal Growth & Design, Vol. 4, No. 5, 2004 933

Figure 7. Change of RT1′′ during temperature induced crystallization processes of 58.3% w/w urea in aqueous solution utilizing three different cooling rates. Also shown in the plot is the solubility of urea in water as measured by ATR FTIR.

Figure 8. Supersaturation profile during a temperatureinduced crystallization process giving the critical supersaturation for spontaneous nucleation. The solution transmittance is plotted to indicate the onset of turbidity and dissolution (concentration: 58.3% w/w urea in aqueous solution; cooling/ heating rate: 0.25 K/min).

plotted against temperature in (see, e.g., Figure 8). Reflecting the high urea concentrations used in the experiments, the “high concentration” exponential functions were used to calculate the concentrations of the SR solutions (see Figure 4). From this analysis, the critical supersaturation values for spontaneous nucleation were found to be 1.05, 1.10, and 1.14 for the three cooling rates of 0.1, 0.25, and 0.33 K/min, respectively. The crystallization and dissolution experiments for the higher concentrated urea solutions (180 g of urea/ 100 g of water) were also monitored by ATR FTIR spectroscopy. Here again, it was found that the onset of crystallization was detected via ATR FTIR prior to the onset of turbidity. However, the supersaturation was found to be underestimated by about 0.12 for all runs conducted for this solution concentration reflecting difficulties in extrapolating the data analysis beyond its calibration range. 3.2. Crystallization of Urea from 20%(w/w) Water and 80%(w/w) Methanol Solution. 3.2.1 Metastable Zone Width Determination. Examination of the measured MSZWs as a function of cooling/heating rates (Figure 9) revealed the runs with the lowest concentra-

934

Crystal Growth & Design, Vol. 4, No. 5, 2004

Figure 9. Average dissolution and crystallization temperatures urea in w-m as a function of the cooling rate. From this, the MSZW can be extrapolated for a cooling rate of 0 K/min.

tion of urea follow the expected trend that the crystallization temperature increases with decreasing cooling rate. At the higher concentration of 593 g of urea per 1000 g of water-methanol (w-m) the crystallization temperatures are found to be rather constant for all cooling rates. Interestingly, the experiments with the highest concentration of urea (655 g of urea per 1000 g of w-m) exhibit an apparent increase in crystallization temperature with increasing cooling rate. The reason for this effect is not clear at this point. Given that this effect was not observed for the binary system (see Figure 1), it is attractive to relate this to nonideality for the tertiary solution system, but clearly further study is needed to quantify this. The MSZWs extrapolated to a zero cooling rate were found to be 8.88 K for 519 g of urea in 1000 g of w-m, 8.81 K for 593 g of urea in 1000 g of w-m and 11.86 K for 655 g of urea in 1000 g of w-m. Hence, in comparison (section 3.1.1) with the higher solubility binary system (water/urea), the mixed solvent system does not crystallize as easily. It is not clear, at this stage, whether this relates to the solubility (i.e., lower for the mixed solvent) or to the more complexity of a mixed solvent system. 3.2.2. ATR FTIR Spectra of Urea-Water-Methanol Solutions. ATR FTIR spectra of urea in w-m solutions are shown in Figure 10. The spectra show three urea bands (b, c, and f), two methanol bands (a and d), and one water band (e) that clearly change in intensity with concentration of urea. The broad band from 3100 to 3500 cm-1 consists of overlapping water, methanol, and amine bands. Due to this, it was not used for the determination of a calibration parameter. The peaks c and d appeared to be very weak which also excluded them from use for calibration. The urea peak f, utilized for calibration in aqueous solution, is overlapping with two weak methanol bands. After closer observation, it was decided that the methanol peak a and the urea peak b were best used to establish the calibration parameter RT1, as they exhibit the most significant intensity changes with concentration. It was noteworthy that for the urea in the w-m system no significant peak shifts were observed over the whole temperature and concentration range investi-

Groen and Roberts

Figure 10. ATR FTIR spectra of urea w-m solutions compared to the spectrum the pure solvent w-m. The following peaks were observed to change characteristically with concentration of urea: (a) methanol band, (b) urea band, (c) urea band, (d) methanol band, (e) water band, and (f) urea band. Spectra shown here were recorded at 50 °C; spectra at 10, 30, and 70 °C follow the same trend.

Figure 11. Calibration of the transmittance ratio RT1 versus the concentration of urea in w-m solution.

gated. This, perhaps, reflects the much lower degree of hydrogen bonding for the mixed methanol/water, when compared to the pure aqueous solution. 3.2.3. ATR FTIR Calibration for Urea-WaterMethanol Solutions. For the mixed solvent system, the calibration parameter RT1 was defined as follows:

RT1 ) transmittance of methanol band at 2835 cm-1 (4) transmittance of urea band at 1612 cm-1 The peak intensities at fixed wavenumbers were used for calibration and the respective calibration curves are given in Figure 11. The calibration curves can be seen to overlap substantially indicating a small influence of temperature on the ATR FTIR spectra. The data points can be fitted by the curve fitting function RT ) a exp(bcs) previously used1 with a and b being a function of temperature. This is in agreement with the calibration behavior for aqueous urea solutions, where up to a concentration of 30% (w/w) urea in water the data points also followed an exponential trend according to the Lambert-Beer Law. 3.2.4. Solubility of Urea in Water-Methanol Solution. Using the calibration parameter RT1 as

Crystallization of Urea

Figure 12. Solubility data as determined by ATR FTIR spectroscopy for urea in w-m. Also shown are solubility data as determined by turbidometric methods.6

Figure 13. Change of RT1 during a temperature-induced crystallization process. Also shown in the plot is the solubility of urea in w-m as measured by ATR FTIR. The solution transmittance is plotted to indicate the onset of turbidity and dissolution (concentration: 519 g of urea/1000 g w-m; cooling/ heating rate: 0.1 K/min).

defined in eq 4, the solubility of urea in w-m was determined by ATR FTIR spectroscopic measurements (Figure 12). Also given in Figure 13 are the solubility data which were determined from the dissolution temperatures using optical turbidity measurements. It can be seen that, for the temperature range of 30 to 40 °C, the solubility as determined using RT1 compared to the equilibrium concentrations of 34.2, 37.2, and 39.6% (w/ w) differ from those determined by turbidity measurements, reflecting again the greater sensitivity to concentration variation of the FTIR technique compared to the turbidometric technique. 3.2.5. Supersaturation Measurements during Urea Crystallization from Water-Methanol Solutions. Supersaturation profiles during the crystallization of urea from w-m were successfully recorded online. Figure 14 shows a typical profile of the raw data RT1 as recorded by ATR FTIR spectroscopy, revealing that the onset of crystallization was detected prior to that observed using the turbidometric probe. This result also agrees with the results observed for the aqueous urea solutions (see Section 3.1.5). This has been observed for all crystallization experiments for urea from w-m solutions and aqueous solutions. The reason for the slight difference of RT1 between the heating and cooling curve in the undersaturated region can be explained by the fact that not all crystals

Crystal Growth & Design, Vol. 4, No. 5, 2004 935

Figure 14. Change of RT1 during crystallization and dissolution processes of differently concentrated urea w-m solutions. Also shown in the plot is the solubility of urea in w-m as measured by ATR FTIR. The MSZW can be extracted from the ATR FTIR data as indicated by the dotted line (cooling/heating rate: 0.1 K/min).

dissolved immediately upon heating. For example, it is possible that some crystals from time to time might get trapped in the gap of the transmission turbidity probe and only dissolve after a given time at a high temperature within the undersaturated region. Figure 13 gives the RT1 profiles during the crystallization experiments at a cooling rate of 0.1 K/min for the three different concentrated solutions of urea. This represents a solubility-supersolubility diagram in terms of RT data, from which the (rate-dependent) MSZW can be extracted. This shows that, if no explicit concentration-based supersaturation value is required, calibration of the ATR FTIR data is not directly needed, simplifying the application of the technique to a great extent. In this case, the supersaturation could be simply expressed by the following ratio:

SRT )

RT1 RT1*

(5.5)

with RT1 being the current relative transmittance and RT1* being the value for a saturated solution at the respective temperature. It should be noted that for ideal behavior according to Lambert-Beer Law, the supersaturation as calculated from an ATR FTIR absorbance ratio (RA) equals the supersaturation calculated from a concentration ratio making the calibration procedure unnecessary. This is correct, if RA and the respective concentration are related linearly with an y-axis intercept of zero. To validate the use of the calibration curves for RT1 (Figure 11) at the edge of the recorded concentration range, the supersaturation profiles during the crystallization runs were extracted (Figure 15). These data clearly illustrates that ATR FTIR spectroscopy gives reliable and predictable results for concentration and therefore supersaturation within the calibration range, even though the urea peak utilized for determination of RT1 substantially overlaps with a solvent band. However, it can be seen that, for the crystallization run of 519 g/1000 g of water, the calibration model reveals that the solution appears to become undersaturated

936

Crystal Growth & Design, Vol. 4, No. 5, 2004

Groen and Roberts

changes in FTIR peak position associated with strong hydrogen bonding in the highly concentrated aqueous urea solutions, no similar evidence for solute structuring has been observed in this work either for aqueous or for w-m solutions. Finally, it should be noted that FTIR calibrations carried out in this work have been based on a univariant (peak intensity) methodology. Further advances in the application of in-process ATR FTIR spectroscopy can be expected through the use of multivariant analysis via the use of chemometric-based calibration using a wider range of FTIR spectral information that has been addressed at present. This work will be reported in future papers. Figure 15. Supersaturation profiles during crystallization and dissolution processes of differently concentrated urea w-m solutions (cooling/heating rate: 0.1 K/min) as determined using the calibration curves from Figure 13. The values for the critical supersaturation for the respective runs are extracted given.

before reaching the saturation temperature during the heating cycle. The reason for this behavior is not completely clear at this stage but could be consistent with the presence of temperature concentration gradients within the 400 mL batch crystallizer vessel. 4. Conclusions ATR FTIR has been successfully shown to be a useful in-process tool for monitoring the supersaturation of organic systems in aqueous or organic solutions. The experiments on urea solutions demonstrate that ATR FTIR spectroscopy can be applied to monitor supersaturation of systems where solute and solvent bands overlap substantially as long as they display a systematic change in intensity with changing concentration and are sufficiently intense to avoid corruption from background noise. It has been shown that the defined transmittance ratios can determine the onset of crystallization and dissolution reliably and provide a valuable parameter from which to deduce the supersaturation of a system. For aqueous urea solutions, a peak shift was observed due to the high degree of hydrogen bonding. The peak shift itself, however, could not be identified to monitor crystallization phenomena as suggested by other authors.7 However, it was found that using the intensity of the peak at a set wavenumber on one site of its maximum provides a parameter exhibiting greater concentration dependency and being therefore more suitable and straightforward for supersaturation determination. In-process ATR FTIR spectroscopy was found to be effective in the detection of the onset of crystallization associated with the depletion of the solution concentration prior to the onset of optical turbidity, hence illustrating the sensitivity of the ATR technique for these kind of measurements. This provides supportive evidence for the observation that the onset of turbidity prior to the crystallization of citric acid1 is consistent with liquid-phase separation solution structuring prior to the onset of crystallization. Apart from the slight

Acknowledgment. This work, which formed part of the Ph.D. thesis of one of us (H.G.),11 has been carried out as part of the Chemicals Behaving Badly initiative, a collaborative project funded by EPSRC Grant GR/L/ 68797 together with industrial support from Astra Charnwood, BASF, GlaxoWellcome, ICI, Malvern Instruments Ltd., Pfizer, SmithKlineBeecham, and Zeneca. We gratefully acknowledge these sponsors and all members of this academic/industrial team, notably, industrial coordinator L. J. Ford. We are also grateful to one of the group’s project students (A. D. McDowall) for his help in collecting some of the data presented in this paper. References (1) Groen, H.; Roberts, K. J. J. Phys. Chem. B 2001, 105, 10723-10730. (2) Baker, M.; Cao, Z.; Dale, D.; Erk, P.; Ford, L. J.,; Groen, H.; Latham, D.; Hammond, R. B.; Lai, X.; Liang, K.; Merrifield, D.; Mougin, P.; Oliver, R.; Roberts, D.; Roberts, K. J.; Savelli, N.; Thomas, A.; White, G.; Wilkinson, D.; Wood, W. Mol. Liq. Cryst. 2001, 356, 273-287. (3) Roberts, K. J. Chem. Eng. Trans. 2002, 1, 773-778. (4) Dunuwila, D. D.; Carroll, L. B.; Berglund, K. A. J. Cryst. Growth 1994, 137, 561-568. (5) Dunuwila, D. D.; Carroll, L. B.; Berglund, K. A. J. Cryst. Growth 1997, 179, 185-193. (6) Dunuwila, D. D.; Berglund, K. A. Org. Proc. Res. Dev. 1997, 1, 350-354. (7) Uusi-Penttila¨, M. S.; Berglund, K. A. Proceedings of the 4th International Workshop on Crystal Growth of Organic Materials, Shaker Verlag: Aachen, 1997; pp 245-252. (8) Groen, H.; Roberts, K. J. Proceedings of the 14th International Symposium on Industrial Crystallization, Paper 77; IChemE: Cambridge, 1999. (9) Lewiner, F.; Fe´votte, G.; Klein, J. P.; Pfefer, G. Proceedings of the 14th International Symposium on Industrial Crystallization, Paper 47; IChemE: Cambridge, 1999. (10) Wang, F.; Berglund, K. A. Ind. Chem. Res. 2000, 39, 21012104. (11) Lewiner, F.; Klein, J. P.; Puel, F.; Fe´votte, G. Chem. Eng. Sci. 2001, 56, 2069-2084. (12) Groen, H.; Borissova, A.; Roberts, K. J. Ind. Eng. Chem. Res. 2003, 42, 198-206. (13) Groen, H.; Mougin, P.; Thomas, A.; White, G.; Wilkinson, D.; Hammond, R. B.; Lai, X.; Roberts, K. J. Ind. Eng. Chem. Res. 2003, 42, 4888-4898. (14) Groen, H. On-Line Monitoring and Control of Supersaturation in Batch Crystallisers for Organic Fine Chemical Products using ATR-FTIR Spectroscopy, Ph.D. Thesis, Heriot-Watt University, 2001. (15) Pink, L. A.; Kelly, M. A. J. Am. Chem. Soc. 1925, 47, 2170-2172.

CG030038Y